Timing edge placement is often a critical parameter for high performance semiconductor testers. Having the ability to place the rising and/or falling edge of a test signal within a few picoseconds of a desired point in time may mean the difference in passing or failing large numbers of semiconductor devices under test.
Conventional timing generators that produce high accuracy timing signals are often employed in CMOS integrated circuits. CMOS technology provides relatively good performance at very low cost. However, CMOS ICs are often susceptible to temperature and other conditions that affect the performance of the circuit. To counter this, many CMOS timing generators employ sophisticated compensation techniques to minimize changes in delay.
With reference to FIG. 1, a conventional CMOS timing generator 10 that provides for temperature compensation typically includes a plurality of delay elements D1–DN coupled together to form a delay line. Each of the delay element outputs serve as timing selection inputs to a timing signal selector (not shown). The same outputs are also used for a delay compensation scheme. A compensation multiplexer 12 is employed, that receives the delay outputs, and provides an output to a phase detector 14, where it is compared to a reference signal Vref to determine any phase difference. A compensation voltage is then generated in response to the magnitude of any phase difference, and fed to a charge pump or voltage-to-current converter 16. The current generated by the converter is provided as a reference bias current to a bias current fanout circuit 18 that duplicates and distributes bias current to the delay elements to control the delay.
To fanout, or distribute the bias current to the various delay elements, a current mirror circuit scheme is typically employed. As shown in FIG. 2, a conventional current mirror includes a reference current source 19 coupled to a current source transistor QS to generate the same current through a first mirror transistor QFM1. The mirror transistor's gate is tied to its drain, with its source terminal coupled to the supply voltage VDD. A plurality of mirroring transistors QFM2–QFMN are disposed in parallel, each with its gate tied to the first mirror transistor gate, and source terminals tied to VDD. 
While this configuration works well for its intended applications, the current through each mirror transistor may be susceptible to noise components acting on the VDD bus. Even a relatively small change of 100 millivolts can cause a corresponding change in the bias current through each mirror, correspondingly affecting the level of current provided therefrom. As voltage levels on integrated circuits continue to decrease, this problem becomes more pronounced.
What is needed and currently unavailable is a compensation circuit for use with current mirror circuits to minimize changes in current attributable to supply voltage noise. The current mirror compensation circuit described herein satisfies this need.